Communication between physicians and patients is often problematic [15-17] and can have negative effects on outcomes [18, 19]. Communication about HIV risk is further complicated by prejudice and fear among health professionals [20-23] and by complex family, sociocultural, and ethical issues [24]. Although educational programs can improve physicians' general communication skills [16, 25, 26] and patient outcomes [18, 27], improving physicians' HIV risk-assessment practices has been difficult [7, 13, 28, 29]. This may be because recommendations on ways to assess HIV risk have been based on theoretical models rather than on data empirically derived from direct observation of physicians' interactions with actual patients. To describe the structure of HIV-related discussions, characterize effective and efficient communication, and identify common difficulties, we conducted an exploratory study by analyzing videotapes of patient–physician encounters in community-based settings.

Methods

Top Methods Results Discussion Author & Article Info References

Physician and Patient Recruitment

We invited all inner-city, community-based general internists and family physicians and a convenience sample of suburban, rural, and urban general internists and family physicians in the greater Rochester, New York, area to participate in a study of "communication about prevention and risk behaviors." In return for participation, physicians received an educational feedback letter, a copy of a videotape and transcript for further review, and continuing medical education credits. We recruited physicians until we had at least 25 videotaped encounters that were eligible for analysis.

Patients 18 to 45 years of age who visited consenting physicians' offices were asked to participate until either two videotapes eligible for review or a total of five videotapes per physician were obtained. Visits were considered eligible for videotaping if patients had presented for a routine physical examination or indicated concern about, risks for, or previous testing for HIV infection on a 10-item written questionnaire that they completed in the examination room before seeing the physician.

Data Collection

Data were collected between May 1995 and September 1996. After we obtained consent from both physician and patient, we collected demographic information and videotaped the office visit. A research assistant reviewed the videotape with the patient. The patient was instructed to stop the tape whenever he or she had a comment about the process of the interview, with particular attention to discussions about HIV [30]. In addition, the research assistant stopped the tape periodically and after any discussion about HIV if the patient had not already done so. All comments were audiotaped. If HIV-related concerns were expressed during the visit or the patient review, the videotape qualified for analysis and was then reviewed in a similar manner by the physician, who did not have access to the patient's comments. At the end of each review, a brief semi-structured interview was administered to evaluate physician and patient concerns not expressed in the office visit and perceived communication barriers to the assessment of HIV risk behaviors. Videotapes and audiotapes were transcribed and edited into a single transcript for each encounter.

Data Analysis

Members of the research team reviewed the first 10 videotapes and transcripts to develop a coding scheme. We condensed themes that arose during the review process into codes that described observable behaviors (behavioral codes) and comments made by physicians, patients or the research team (interpretive codes). We paid particular attention to specific interactions that seemed to inhibit or facilitate risk assessment. We calculated an initial interrater reliability score to guide further development of the coding scheme. We then analyzed additional interviews and added, defined, combined, and eliminated codes until we found no new codes (saturation [31]). Subsequent assignment of codes was done by consensus. A code was assigned if the behavior or comment was present in any part of the tapes.

Next, we created a five-point scale to describe the overall depth of inquiry about HIV during the office visit (Table 1). Level of inquiry scores were assigned by two reviewers who were blinded to the coding categories, and interrater reliability was calculated. Descriptive statistics and simple correlations were calculated to describe demographic data and show relations among demographic variables, interview behaviors, and levels of inquiry.

We were interested not only in behaviors and overall depth of discussion but also in the process, or flow, of the discussion. First, we looked informally at the sequence of utterances in all 26 interviews to identify common patterns in the HIV-related discussions. We then formally analyzed each interview to produce a model for conversational flow.

Results

Top Methods Results Discussion Author & Article Info References

Demographic characteristics of the physicians are shown in Table 2. We videotaped 78 office visits. In 31 cases, HIV was discussed during the office visit or the patient review session. Four patients did not review the tapes despite several reminders, and equipment failure precluded the use of one patient review. Thus, 26 patient encounters were eligible for analysis. The mean age of the 26 patients was 38 years. Sixteen patients were women, 9 were African American, and 6 were Hispanic.

The qualifying encounters and review comments consisted of 668 minutes of office visit dialogue and 973 pages of transcript. The median length of HIV-related discussion was 108 seconds (mean ±SD, 149 ± 161 seconds; range, 0 to 557 seconds). Frequencies of codes are shown in Table 3. Eleven interviews that had not been previously analyzed by the research team were dual-coded with an interrater agreement of 0.86 and an intraclass correlation of 0.67. Items with poor correlation and agreement were dropped.

Interrater reliability (intraclass correlation) for the level of inquiry scale (Table 1) was 0.73. We defined a comprehensive evaluation of a patient at risk for or concerned about HIV as level 3 or 4. Comprehensive interviews took 3 minutes longer than incomplete HIV risk assessments (P < 0.001).

Conversational Sequences

We used consensual validation to examine the conversational sequences in detail and create a model that describes patterns of interaction between physicians and patients during HIV-related discussions (Figure 1). The model has three critical branch points. The first is at the opening of the HIV-related discussion. At this point, some physicians either provided a rationale for bringing up the topic (such as, "I ask all of my patients about HIV") or made explicit the context for the discussion ("You mentioned that you had an HIV test for insurance purposes and that you were nervous about it. Could you tell me why you were nervous? Do you think you might be at risk for HIV?"). Other physicians made a more abrupt or decontextualized transition (Table 4).

View larger version (31K): [in this window] [in a new window]   Figure 1. HIV-related discussions.

The second branch point, handling of awkward moments and problematic language, was the most richly illustrated in our sample. The segments were problematic because HIV-related discussions characteristically were abandoned at an awkward moment, and neither patients nor physicians reintroduced the topic. We defined an awkward moment to be an event during a clinical encounter when either the physician or patient showed enough uneasiness, embarrassment, distraction, or nervousness to disrupt the flow of conversation. In our initial coding, we noted that the ability to sustain conversation and not change the topic during an awkward moment often made a difference in the amount of clinical information that the physician was able to collect. When we restudied the transcripts, we found awkward moments in all interviews, regardless of the depth of inquiry, and defined them further. Awkward moments included the use of euphemisms, inappropriate humor, leading questions, statements of ambivalence, abrupt transitions, and uncomfortable silences.

We found a similar pattern with problematic language, language that obscured the elicitation of important information. All interviews contained at least one example of problematic language, but the clinician's handling of the language varied. Types of problematic language included vague statements, misstatements, judgmental language, and loaded questions. We also encountered fractured language, a particular type of problematic language characterized by multiple hesitations and reformulations.

The third branch point, seeking the patient's perspective, pertains to physicians who achieved at least a level 3 assessment (Table 1). Although some physicians terminated the interview after gathering information about behavior, those who were rated at level 4 sought an understanding of the patient's attributions, motivations, and emotions.

Opening Moments of HIV-Related Discussions

Four of the 22 HIV-related discussions were initiated by patients. Patients commented that they expected physicians to initiate HIV-related discussions. However, when physicians did initiate discussions, the transitions were often abrupt (Table 4): In only 4 of the interviews did the physician explain why he or she was eliciting information about HIV risk, and only 3 discussions began with an open-ended inquiry. Nonverbal barriers to effective communication were frequent. For example, physicians were observed to shake their heads "no," avoid eye contact, or turn their back to the patient while asking about sexual behavior and to initiate discussions about HIV during a genital examination.

Ineffective Handling of Awkward Moments

In most of the interviews, the physician avoided discussion of HIV risk by changing the topic, not pursuing patients' cues, or ignoring a stated concern (Table 3). For example, the awkward moment presented in Table 4 occurred when the patient disclosed that her partner does not use condoms all of the time. Instead of pursuing what may have been an uncomfortable discussion about sexual practices, the physician abruptly changed the topic to smoking cessation without returning to HIV risk assessment.

Unrepaired Conversations: Fractured Language and a Conspiracy of Vagueness

Our informal observation that language used during discussions about HIV was often less coherent than that used during other parts of the visit led us to examine HIV discussions in detail. Fractured language was noted by patients, physicians, and the research team to be a sign of anxiety; this type of language was rarely observed in discussions of less highly charged topics. Table 4 gives part of a fractured prologue during which the physician attempted to introduce the topic of sexuality. This portion of the discussion includes multiple hesitations and reformulations, whereas the physician's language in other parts of the encounter was not fractured. The physician's and patient's comments on this segment indicated that they were both nervous. Although HIV risk was discussed further after this sequence, the discussion was characterized by vague statements that were not clarified and leading questions. The HIV risk discussion was long (310 seconds) considering the limited information (level 2) that was gathered.

A conspiracy of vagueness refers to a pattern of vague questions or statements between physician and patient that are never clarified, followed by a change of topic without further characterization of HIV risk behaviors (Table 4). In those cases, presuppositions and incomplete information guided further interventions, precluding complete characterization of the patient's behaviors.

Effective Handling of Awkward Moments and Use of Problematic Language

In comprehensive discussions, physicians were able to sustain discussion about HIV risk despite awkwardness and return to the topic after a digression. Empathy, laughter, and digression seemed to be effective methods of dissipating anxiety. Physicians clarified vague statements made by themselves or the patient, corrected fractured language and misstatements, and returned to the topic to inquire further about HIV risk. Clarifying statements lessened the impact of leading questions. These kinds of conversational repair and reconstruction [32-34] helped physicians to obtain more detailed information and bring the conversation to a satisfactory close. Table 5 shows examples of such conversational barriers as leading questions, vague statements, and awkward moments and gives several methods of overcoming these barriers, such as remaining focused, correcting misunderstandings, using empathy, returning to the topic after a digression, and using closed-ended questions to clarify behavior further.

Eliciting the Patient's Perspective

The four interviews that used a patient-centered approach to get to the heart of the problem took a mean of 1 minute less than level 3 interviews that were otherwise complete. Patient-centered physicians used such phrases as "What things do you worry about?", which often led to further disclosure of potentially stigmatizing information. For example, one patient disclosed her shame and concerns about infections related to anal intercourse. In that case, while exploring the patient's fear and shame, the physician clarified her risk behaviors and achieved an empathic understanding of the patient that could not have been elicited by using a more closed-ended interviewing style. Additional examples of empathic communication are given in Table 5.

Patient and Physician Reflections: Barriers to Successful Discussion of HIV Risk

Physician and patient comments on their videotapes gave clues as to why certain conversations failed. Physicians stopped the tape frequently to comment on the interviews. Physicians were generally surprised at their own discomfort at awkward moments. They offered anxiety, identification with the patient, lack of knowledge, and economic disincentives as explanations for their actions. Some physicians were visibly upset by and extremely critical of their own communication behavior during the review process. In only one case did a physician indicate that HIV risk had been evaluated in a previous visit.

The following are a physician's comments made after he reviewed an interview in which the discussion was truncated. He said that he had chosen a poor time to initiate the discussion and that he avoided discussing HIV because he identified with the patient.

"That's just obviously my discomfort because I rushed through: "Yeah, by the way, do you want to talk about that? Okay, great, I'm glad you don't." That issue [is difficult to] deal with not [because] of knowledge, but attitude. And so, is it that I'm dealing with a man that I need to talk about an issue of sexuality with? I don't know. Is it that I'm an hour behind in this and I don't want to do it? Or is it a topic that I want to talk about at a different time?"

In contrast, during the same interview, the physician discussed another potentially challenging topic (depression) fully. Other physicians offered explanations for their actions, such as lack of time, lack of expertise in HIV, multiple competing concerns, low self-efficacy, desire not to know a patient's HIV status, desire not to refer the patient, and structural constraints in the health care system.

Physician and patient comments on one interview provided an example of projection and complex disagreement [30]. Both physician and patient stopped the tape within 40 seconds of each other at the beginning of their discussion about sexual behavior. Each person recognized that anxiety was present but implicated the other as having been nervous.

Patients stopped the tape infrequently and often did not comment on communication difficulties. However, patients expressed a desire to discuss HIV with their physician and the expectation that the physician would initiate the conversation.

Discussion

Top Methods Results Discussion Author & Article Info References

The outcome of HIV-related discussions was substantially influenced by the manner in which the physician introduced the topic, handled awkward moments, and dealt with problematic language and the extent to which the physician sought the patient's perspective.

Patients in our study wanted their physicians to introduce the topic of HIV, but few physicians had a comfortable standard question or statement with which to introduce the topic. In their reviews, physicians often spoke of the lack of a natural context for these discussions, such as HIV-related symptoms or concerns. Although this observation should be confirmed in larger studies, we suggest that educational programs to improve physician assessment of HIV risk should offer sample transitions or rationales for raising the topic and should help practitioners use initial open-ended questions. An opening statement, such as "This is the era of AIDS, and I ask all my patients if they are concerned about HIV. What concerns do you have?" can alleviate patient fears and anxiety, create a collaborative relationship, and help patients disclose their concerns more freely.

Another factor in successfully discussing sensitive topics is a physical environment that is perceived as psychologically safe. In that regard, we noted that some physicians initiated HIV-related discussions during the physical examination. We know of no data that support this practice. Although it seems innocuous during some routine inquiries, in the area of the sexual history, the patient's feelings of vulnerability and fear are likely to be lessened if the patient is fully dressed.

Once discussions were initiated, few physicians asked enough specific questions to assess risk behaviors that could be modified. Physicians frequently terminated discussions about HIV risk prematurely or did not return to the topic after a digression. Clinicians would benefit from learning ways of dealing with embarrassment and ambivalence about exploring the sexual behaviors of their patients. By recognizing the inevitability of miscommunication, physicians can learn ways to manage and repair difficult conversations [32-36]. These skills include negotiating a range of awkward moments; clarifying problematic language; and attending to the patient's perspective, fears and expectations. Simple clarifying statements, such as "I'm not sure what you meant when you said ... ," or redirecting statements, such as "Would you mind if I asked another question about HIV?" can help overcome some of the barriers. Physicians who learn to persevere despite these difficulties will be rewarded with more information and more opportunities to counsel patients about changes in behavior.

We were impressed by the degree to which some practitioners' feelings of ineffectiveness, identification with patients, and strong emotions interfered with their ability to assess HIV risk. These responses underscore the emotional difficulties that physicians have in dealing with HIV and risk behaviors [20, 37]. Physicians would benefit from taking advantage of opportunities to become more aware of their internal reactions to patients, including embarrassment and ambivalence about inquiring into sexual behavior, and finding ways to deal with those feelings more effectively [38].

Comprehensive assessment of patients with risk behaviors or concerns about HIV required an average of 3 minutes longer than incomplete assessments. Consistent with previous research [39], assessments that incorporated an understanding of the patient's perspective (level 4), which is critical to behavioral change [40-42], took no longer than assessments that were more physician directed (level 3). Clearly, not all patients will require a comprehensive assessment; a few focused screening questions may suffice for patients whose answers to initial questions indicate that they are not currently at risk. However, because patients do not often bring up the topic, the physician must always remain alert to cues, correct miscommunications, and keep the door open for further discussion. Care of HIV-infected patients is expensive, and cost–benefit analysis would probably justify the additional time that clinicians may have to spend to assess and modify risk for HIV infection.

Our sample included four physicians who had fellowship training in communication skills. Our findings show that even among physicians who have had advanced training in patient–physician communication, significant barriers remain in the domain of HIV risk assessment. In addition to teaching generic interviewing skills, educational programs should emphasize competency in HIV risk assessment, including the opportunity to practice with real or standardized patients. We believe that these skills may apply to a wide variety of sensitive topics, not just HIV.

Strengths of our study include the use of descriptive qualitative methods [31, 43, 44] to capture the complexities, ambiguities, and contexts of discourse that are missed when transcribed utterances are simply counted and classified. We chose this approach because research indicates that the outcomes of clinical encounters may relate less to the number of utterances that fall into specific a priori observer-defined codes than to the meaning attributed to the same dialogue by the physician and patient [45]. Using physicians and patients as experts to interpret their own actions and feelings offers insight into the interview and can uncover unexpected barriers, such as lack of a simple opening statement [46]. Purposeful sampling [47] selected for encounters in which HIV-related discussions occurred. Consistent with other studies [48, 49], our physicians reported that videotaping did not change their communication styles.

Our study has several limitations. We successfully recruited only physicians who were willing to participate in an intrusive examination of their communication about difficult topics and were willing to spend 1 to 2 hours reviewing the videotapes of their interviews. Although a random sample would have been expensive and impractical, a larger sample of physicians and patients may have yielded different patterns of interaction and a different set of codes. In addition, we could not evaluate the accuracy of the assessments beyond comments made by patients during the review process.

We hope that our study will generate momentum to improve physicians' communication skills, particularly about highly charged topics. Because even highly trained physicians lack these skills, attention should be focused on improving communication for HIV risk assessment. Videotape review helps physicians become aware of miscommunications and enhances self-awareness [50]. However, given economic and time pressures, practicing physicians are unlikely to participate in educational programs if health systems do not support their efforts. We have noted that medical students and residents show improved skills after participating in HIV-related educational programs; the outcomes of these interventions should be the subject of further research. Meanwhile, we hope that physicians will be less hesitant to open focused inquiries into HIV risk and other sensitive topics with their patients.

Dr. Morse: General Medicine Unit, Rochester General Hospital, 1425 Portland Avenue, Rochester, NY 14621.

Drs. Frankel and Beckman: Department of Internal Medicine, Highland Hospital, 1000 South Avenue, Rochester, NY 14620.